3D Urban Voxel Semantic Classification Dataset and AI Experiment Results for 5G Network Planning.

This dataset contains research data generated during the development and evaluation of artificial intelligence methods for semantic classification of three-dimensional urban environments, with applications in spatial analysis and 5G network planning. The dataset includes voxel-based representations of urban scenes from Berlin, Bydgoszcz, London, Los Angeles, Maceió, Manchester, Nagoya, Paris, Rome, and Stockholm. The voxel representation uses a spatial resolution of 0.5 m. The published data comprise: Voxel-level spatial attributes, RGB values, and semantic classification information. Spatial tile identifiers, metadata, and semantic class definitions. AI model training experiments, computational tasks, configurations, and training history. Model evaluation results, performance metrics, and confusion matrices. Voxel-level predictions from experiments E0–E3. A data dictionary describing the structure, types, and meaning of individual fields. The dataset consists of 24 original CSV files, provided individually to facilitate direct access, independent analysis, and reuse of the research data. A separate README document describes the dataset structure, file relationships, and interpretation of the data. Some source tables contain empty fields where a parameter was unavailable or not applicable; these should not be interpreted as numerical zeros. The data were generated by Blare Technologies Sp. z o.o. within the research and development project: "Inteligentny system do analizy i optymalizacji planów sieci 5G z wykorzystaniem sieci neuronowych (głębokiego uczenia maszynowego) oraz algorytmów eksperckich" Project number: RPPM.01.01.01-22-0028/18-07. The dataset is made publicly available to support reproducibility, further research, and the development of AI-assisted methods for three-dimensional spatial analysis and infrastructure planning.